Descriptive statistics, basic visualizations, variable descriptions, and variable distributions
Descriptive Statistics
Basic Visualizations
The impetus for this study was the rapid increase in governance sector aid after the end of the Cold War, which has been widespread and has made up a significant portion of donor aid portfolios since the early 2000s. These changes are visualized below in a time series plot of governance sector aid as a percentage of total aid across all Western donors from 1980-2020.
The plots below, which visualize the sectoral composition of aid by donor in ten year increments, demonstrate the differences between donors in the sectors of aid allocation and the increasing complexity of the development aid system, demonstrated by the growing number of donors over time.
The below plot shows total aid by sector by donor for the study period and demonstrates the difference in the quantities of aid that each donor provides to each sector. Japan and Korea dominate economic sector aid allocations, followed by the United States, France, and Germany. The United States far outweighs other countries in its governance sector aid allocations, demonstrating its emphasis on democracy promotion and political liberalization.
Variable unit of measure, descriptions, transformations, and data sources
- Ln Governance Aid: Dyad-Year; Natural log of 1 + governance sector aid per 1000 population; OECD CRS, World Bank WDI
- Ln Economic Aid: Dyad-Year; Natural log of 1 + economic sector aid per 1000 population; OECD CRS
- State Violence (SV): Recipient-Year; Inverted dynamic latent human rights score; Fariss, Kenwick, and Reuning (2020)
- Donor Rights: Donor-Year; Dynamic latent human rights score; Fariss, Kenwick, and Reuning (2020)
- UN Ideal Point Distance: Dyad-Year; Measurement of dissimilarity in United Nations voting record; Bailey, Strezhnev, and Voeten (2017)
- Ln Terror Events: Recipient-Year; Natural log of 1 plus the count of terrorist events within a country; ICEWS
- Ln Exports: Dyad-Year; Natural log of 1 + exports from donor to recipient; World Bank WDI
- State Capacity: Recipient-Year; Government Effectiveness Estimate; World Bank WGI
- Executive Constraints: Recipient-Year; Sum of the legislative and judicial constraints on the executive indices; VDEM
- UN Shaming: Recipient-Year; Count of physical integrity shaming in UNCHR and UNHRC; Demeritt and Conrad (2019)
- Military Alliance: Dyad-Year; 1 if formal military alliance between donor and recipient; Gibler (2008)
- Ln Population: Recipient-Year; Natural log of the recipient country’s population; World Bank WDI
- Ln GDP: Recipient-Year; Natural log of the recipient country’s GDP in 2018 constant USD; World Bank WDI
- Aid Concentration: Recipient-Year; Herfindahl-Hirschman Index calculation of the concentration of aid; AidData and OECD CRS
- Conflict: Recipient-Year; 1 if intrastate or interstate conflict within country; UCDP/PRIO (2021)
- High Dissent: Recipient-Year; 1 if in the top quartile by count of peaceful protests; ICEWS
- Statist Donor: Donor; 1 if statist/traditional public sector donor political economy; Dietrich (2021)
Dependent variable distributions
The dependent variables for economic and governance sector aid commitments present a "corner solution" problem for the analysis. Donors can choose to commit no new aid (zero) or a positive amount of aid. This results in a large proportion of zeros relative to positive values. The untransformed sums of donor-recipient-year commitments by sector also have an overdispersion problem with several large outliers. The overdispersion and outliers problems were easily addressed by taking the natural log of 1 plus the summation of aid per 1000 population (we add 1 because the natural log of 1 is 0, thereby preserving zero commitments and making positive commitments mathematically comparable to zero commitments).
The large number of zeros necessitate an estimator that can deal adequately with a continuous variable that has a corner solution at zero. The Tobit model satisfies this requirement.
The full and truncated (positive) distributions of the dependent variables are shown below:
Prepare data for marginal effects plots
summarize repression_L1_demeaned
return list
global repress_hi=round(r(mean)+r(sd),0.1)
global repress_avg=round(r(mean),0.1)
global repress_lo=round(r(mean)-r(sd),0.1)
summarize idealptdist_L1_demeaned
return list
global idealpt_hi=round(r(mean)+r(sd),0.1)
global idealpt_avg=round(r(mean),0.1)
global idealpt_lo=round(r(mean)-r(sd),0.1)
summarize donorrights_L1_demeaned
return list
global drights_hi=round(r(mean)+r(sd),0.1)
global drights_avg=round(r(mean),0.1)
global drights_lo=round(r(mean)-r(sd),0.1)
summarize xdem_L1_demeaned
return list
global xdem_hi=round(r(mean)+r(sd),0.1)
global xdem_avg=round(r(mean),0.1)
global xdem_lo=round(r(mean)-r(sd),0.1)
summarize statecap_L1_demeaned
return list
global statecap_hi=round(r(mean)+r(sd),0.1)
global statecap_avg=round(r(mean),0.1)
global statecap_lo=round(r(mean)-r(sd),0.1)
summarize lnexport_L1_demeaned
return list
global lnexp_hi=round(r(mean)+r(sd),0.1)
global lnexp_avg=round(r(mean),0.1)
global lnexp_lo=round(r(mean)-r(sd),0.1)
summarize shame_physint_L1_demeaned
return list
global shame_hi=round(r(mean)+r(sd),0.1)
global shame_avg=round(r(mean),0.1)
global shame_lo=round(r(mean)-r(sd),0.1)Run Tobit models for the Economic Aid dependent variable, generate marginal effects plots to visualize continuous-by-continuous interactions with moderating variables, and export output as combined Excel table.
set more off
xtset dyad_id year
************************************
* TOBIT MODELS FOR ECONOMIC AID DV *
************************************
* Model 1: Econ Base
tobit lneconaidpc repression_L1_demeaned donorrights_L1_demeaned idealptdist_L1_demeaned lnexport_L1_demeaned statecap_L1_demeaned xdem_L1_demeaned 1.highterror_L1 1.highdissent_L1 lngdp_L1_demeaned lnpop_L1_demeaned hhi_recip_L1_demeaned statist_donor i.year i.donorcode, ll(0) cl(donorcode)
est sto m1_econ_base
margins, at(repression_L1_demeaned=(-4(.5)4))
marginsplot
graph save Graph "graphs\1_econ_base.gph", replace
* Model 2: Econ Donor Rights
tobit lneconaidpc repression_L1_demeaned donorrights_L1_demeaned idealptdist_L1_demeaned lnexport_L1_demeaned statecap_L1_demeaned xdem_L1_demeaned 1.highterror_L1 1.highdissent_L1 lngdp_L1_demeaned lnpop_L1_demeaned hhi_recip_L1_demeaned statist_donor i.year c.donorrights_L1_demeaned#c.repression_L1_demeaned, ll(0) cl(donorcode)
est sto m2_econ_donorrights
margins, dydx(repression_L1_demeaned) at(donorrights_L1_demeaned=(-4(2)4))
margins, at(repression_L1_demeaned=(-4(1)4) donorrights_L1_demeaned=($drights_lo $drights_avg $drights_hi))
marginsplot, legend(position(6))
graph save Graph "graphs\2_econ_donorrights.gph", replaceThis interaction is not statistically significant, and there is no evidence that donors with stronger respect for human rights in their own country are more likely to use aid as a reward and punishment to promote human rights in other countries. This type of donor is the most likely to prioritize human rights, so that we do not see a relationship here casts doubt on the idea that donors use aid in this manner. The slopes of the +/- 1 standard deviation and mean linear predictions do not significantly differ from each other.
* Model 3: Econ Ideal Point
tobit lneconaidpc repression_L1_demeaned donorrights_L1_demeaned idealptdist_L1_demeaned lnexport_L1_demeaned statecap_L1_demeaned xdem_L1_demeaned 1.highterror_L1 1.highdissent_L1 lngdp_L1_demeaned lnpop_L1_demeaned hhi_recip_L1_demeaned statist_donor c.idealptdist_L1_demeaned#c.repression_L1_demeaned i.year i.donorcode, ll(0) cl(donorcode)
est sto m3_econ_idealpt
margins, dydx(repression_L1_demeaned) at(idealptdist_L1_demeaned=(-1.5(2)3))
margins, dydx(repression_L1_demeaned) at(idealptdist_L1_demeaned=($idealpt_lo $idealpt_avg $idealpt_hi)) pwcompare(effects)
margins, at(repression_L1_demeaned=(-4(1)4) idealptdist_L1_demeaned=($idealpt_lo $idealpt_avg $idealpt_hi))
marginsplot, legend(position(6))
graph save Graph "graphs\3_econ_idealpt.gph", replaceThis interaction is not statistically significant, and there is no evidence that donors give preferential treatment to recipient countries that vote with them in the United Nations, which is a common measure of policy influence and political affinity between countries. The slopes of the +/- 1 standard deviation and mean linear predictions do not significantly differ from each other.
* Model 4: Econ Export
tobit lneconaidpc repression_L1_demeaned donorrights_L1_demeaned idealptdist_L1_demeaned lnexport_L1_demeaned statecap_L1_demeaned xdem_L1_demeaned 1.highterror_L1 1.highdissent_L1 lngdp_L1_demeaned lnpop_L1_demeaned hhi_recip_L1_demeaned statist_donor c.lnexport_L1_demeaned#c.repression_L1_demeaned i.year i.donorcode, ll(0) cl(donorcode)
est sto m4_econ_lnexport
margins, dydx(repression_L1_demeaned) at(lnexport_L1_demeaned=(-14(6)10))
margins, dydx(repression_L1_demeaned) at(lnexport_L1_demeaned=($lnexp_lo $lnexp_avg $lnexp_hi)) pwcompare(effects)
margins, at(repression_L1_demeaned=(-4(1)4) lnexport_L1_demeaned=($lnexp_lo $lnexp_avg $lnexp_hi))
marginsplot, legend(position(6))
graph save Graph "graphs\4_econ_lnexport.gph", replace
This interaction is not statistically significant, and there is no evidence that donors give preferential treatment to their trading partners. The slopes of the +/- 1 standard deviation and mean linear predictions do not significantly differ from each other.
* Model 5: Econ Executive Constraints
tobit lneconaidpc repression_L1_demeaned donorrights_L1_demeaned idealptdist_L1_demeaned lnexport_L1_demeaned statecap_L1_demeaned xdem_L1_demeaned 1.highterror_L1 1.highdissent_L1 lngdp_L1_demeaned lnpop_L1_demeaned hhi_recip_L1_demeaned statist_donor c.xdem_L1_demeaned#c.repression_L1_demeaned i.year i.donorcode, ll(0) cl(donorcode)
est sto m5_econ_execconst
margins, dydx(repression_L1_demeaned) at(xdem_L1_demeaned=(-1(1)1))
margins, dydx(repression_L1_demeaned) at(xdem_L1_demeaned=($xdem_lo $xdem_avg $xdem_hi)) pwcompare(effects)
margins, at(repression_L1_demeaned=(-4(1)4) xdem_L1_demeaned=($xdem_lo $xdem_avg $xdem_hi))
marginsplot, legend(position(6))
graph save Graph "graphs\5_econ_execconst.gph", replaceFinally! There is statistically- and substantively-significant evidence that something moderates the relationship between state violence and economic sector aid.
Legislative and judicial constraints on the executive -- that is, democratic checks and balances that keep countries' leaders from being able to act unilaterally -- condition whether donors choose to punish violations and reward compliance with human rights standards.
But this has nothing to do with selfish donor interests, as the prior literature assumes. On the contrary, the downward sloping black line and crossover effects (the 'bowtie' shape) demonstrate that donors use economic sector aid as both a reward and a punishment with more autocratic states, but not with democracies. Donors selectively use economic aid for coercive influence where leaders in recipient states wield more power and face fewer constraints.
* Model 6: Econ State Capacity
tobit lneconaidpc repression_L1_demeaned donorrights_L1_demeaned idealptdist_L1_demeaned lnexport_L1_demeaned statecap_L1_demeaned xdem_L1_demeaned 1.highterror_L1 1.highdissent_L1 lngdp_L1_demeaned lnpop_L1_demeaned hhi_recip_L1_demeaned statist_donor c.statecap_L1_demeaned#c.repression_L1_demeaned i.year i.donorcode, ll(0) cl(donorcode)
est sto m6_econ_statecap
margins, dydx(repression_L1_demeaned) at(statecap_L1_demeaned=(-2(1)2))
margins, dydx(repression_L1_demeaned) at(statecap_L1_demeaned=($statecap_lo $statecap_avg $statecap_hi)) pwcompare(effects)
margins, at(repression_L1_demeaned=(-4(1)4) statecap_L1_demeaned=($statecap_lo $statecap_avg $statecap_hi))
marginsplot, legend(position(6))
graph save Graph "graphs\6_econ_statecap.gph", replace
* Model 7: Econ High Terrorism
tobit lneconaidpc repression_L1_demeaned donorrights_L1_demeaned idealptdist_L1_demeaned lnexport_L1_demeaned statecap_L1_demeaned xdem_L1_demeaned 1.highterror_L1 1.highdissent_L1 lngdp_L1_demeaned lnpop_L1_demeaned hhi_recip_L1_demeaned statist_donor 1.highterror_L1#c.repression_L1_demeaned i.year i.donorcode, ll(0) cl(donorcode)
est sto m7_econ_terror
margins, dydx(1.highterror_L1) at(repression_L1_demeaned=(-4(1)4)) pwcompare(effects)
margins highterror_L1, at(repression_L1_demeaned=(-4(1)4))
marginsplot, legend(position(6))
graph save Graph "graphs\7_econ_terror.gph", replace* Model 8: Econ Shaming
tobit lneconaidpc repression_L1_demeaned donorrights_L1_demeaned idealptdist_L1_demeaned lnexport_L1_demeaned statecap_L1_demeaned xdem_L1_demeaned 1.highterror_L1 1.highdissent_L1 lngdp_L1_demeaned lnpop_L1_demeaned hhi_recip_L1_demeaned shame_physint_L1_demeaned statist_donor i.year i.donorcode c.shame_physint_L1_demeaned#c.repression_L1_demeaned, ll(0) cl(donorcode)
est sto m8_econ_shaming
margins, dydx(repression_L1_demeaned) at(shame_physint_L1_demeaned=(-.25(.25)1))
margins, dydx(repression_L1_demeaned) at(shame_physint_L1_demeaned=($shame_lo $shame_avg $shame_hi)) pwcompare(effects)
margins, at(repression_L1_demeaned=(-4(1)4) shame_physint_L1_demeaned=($shame_lo $shame_avg $shame_hi))
marginsplot, legend(position(6))
graph save Graph "graphs\8_econ_shaming.gph", replace
*****************************
* TABLE: ECONOMIC SECTOR DV *
*****************************
global econtobit m1_econ_base m2_econ_donorrights m3_econ_idealpt m4_econ_lnexport m5_econ_execconst m6_econ_statecap m7_econ_terror m8_econ_shaming
outreg2 [${econtobit}] using "tables\econ_table.xls", label replace excelThe Governance Aid dependent variable
Run Tobit models for the Governance Aid dependent variable, generate marginal effects plots to visualize continuous-by-continuous interactions with moderating variables, and export output as combined Excel table.
* Model 9: Gov Base
tobit lngovaidpc repression_L1_demeaned donorrights_L1_demeaned idealptdist_L1_demeaned lnexport_L1_demeaned statecap_L1_demeaned xdem_L1_demeaned 1.highterror_L1 1.highdissent_L1 lngdp_L1_demeaned lnpop_L1_demeaned hhi_recip_L1_demeaned statist_donor i.year i.donorcode, ll(0) cl(donorcode)
est sto m9_gov_base
margins, at(repression_L1_demeaned=(-4(.5)4))
marginsplot
graph save Graph "graphs\9_gov_base.gph", replaceHere we see a much stronger relationship than was reflected in the economic aid dependent variable. In this case, there is a strong positive relationship between levels of state violence and aid to projects that would improve governance, support free and fair elections, and address post-conflict violence.
* Model 10: Gov Donor Rights
tobit lngovaidpc repression_L1_demeaned donorrights_L1_demeaned idealptdist_L1_demeaned lnexport_L1_demeaned statecap_L1_demeaned xdem_L1_demeaned 1.highterror_L1 1.highdissent_L1 lngdp_L1_demeaned lnpop_L1_demeaned hhi_recip_L1_demeaned statist_donor i.year c.donorrights_L1_demeaned#c.repression_L1_demeaned, ll(0) cl(donorcode)
est sto m10_gov_donorrights
margins, dydx(repression_L1_demeaned) at(donorrights_L1_demeaned=(-4(2)4))
margins, dydx(repression_L1_demeaned) at(donorrights_L1_demeaned=($drights_lo $drights_avg $drights_hi)) pwcompare(effects)
margins, at(repression_L1_demeaned=(-4(1)4) donorrights_L1_demeaned=($drights_lo $drights_avg $drights_hi))
marginsplot, legend(position(6))
graph save Graph "graphs\10_gov_donorrights.gph", replaceThe upward sloping blue line demonstrates that donors with stronger domestic respect for human rights exhibit stronger catalytic responses to state violence than other donors. This correlation suggests that the donors that care most about promoting human rights abroad are more likely to adopt catalytic strategies.
* Model 11: Gov Ideal Point
tobit lngovaidpc repression_L1_demeaned donorrights_L1_demeaned idealptdist_L1_demeaned lnexport_L1_demeaned statecap_L1_demeaned xdem_L1_demeaned 1.highterror_L1 1.highdissent_L1 lngdp_L1_demeaned lnpop_L1_demeaned hhi_recip_L1_demeaned statist_donor c.idealptdist_L1_demeaned#c.repression_L1_demeaned i.year i.donorcode, ll(0) cl(donorcode)
est sto m11_gov_idealpt
margins, dydx(repression_L1_demeaned) at(idealptdist_L1_demeaned=(-1.5(2)3))
margins, dydx(repression_L1_demeaned) at(idealptdist_L1_demeaned=($idealpt_lo $idealpt_avg $idealpt_hi)) pwcompare(effects)
margins, at(repression_L1_demeaned=(-4(1)4) idealptdist_L1_demeaned=($idealpt_lo $idealpt_avg $idealpt_hi))
marginsplot, legend(position(6))
graph save Graph "graphs\11_gov_idealpt.gph", replace* Model 12: Gov Export
tobit lngovaidpc repression_L1_demeaned donorrights_L1_demeaned idealptdist_L1_demeaned lnexport_L1_demeaned statecap_L1_demeaned xdem_L1_demeaned 1.highterror_L1 1.highdissent_L1 lngdp_L1_demeaned lnpop_L1_demeaned hhi_recip_L1_demeaned statist_donor c.lnexport_L1_demeaned#c.repression_L1_demeaned i.year i.donorcode, ll(0) cl(donorcode)
est sto m12_gov_export
margins, dydx(repression_L1_demeaned) at(lnexport_L1_demeaned=(-14(6)10))
margins, dydx(repression_L1_demeaned) at(lnexport_L1_demeaned=($lnexp_lo $lnexp_avg $lnexp_hi)) pwcompare(effects)
margins, at(repression_L1_demeaned=(-4(1)4) lnexport_L1_demeaned=($lnexp_lo $lnexp_avg $lnexp_hi))
marginsplot, legend(position(6))
graph save Graph "graphs\12_gov_export.gph", replace
* Model 13: Gov Executive Constraints
tobit lngovaidpc repression_L1_demeaned donorrights_L1_demeaned idealptdist_L1_demeaned lnexport_L1_demeaned statecap_L1_demeaned xdem_L1_demeaned 1.highterror_L1 1.highdissent_L1 lngdp_L1_demeaned lnpop_L1_demeaned hhi_recip_L1_demeaned statist_donor c.xdem_L1_demeaned#c.repression_L1_demeaned i.year i.donorcode, ll(0) cl(donorcode)
est sto m13_gov_execconst
margins, dydx(repression_L1_demeaned) at(xdem_L1_demeaned=(-1(1)1))
margins, dydx(repression_L1_demeaned) at(xdem_L1_demeaned=($xdem_lo $xdem_avg $xdem_hi)) pwcompare(effects)
margins, at(repression_L1_demeaned=(-4(1)4) xdem_L1_demeaned=($xdem_lo $xdem_avg $xdem_hi))
marginsplot, legend(position(6))
graph save Graph "graphs\13_gov_execconst.gph", replace* Model 14: Gov State Capacity
tobit lngovaidpc repression_L1_demeaned donorrights_L1_demeaned idealptdist_L1_demeaned lnexport_L1_demeaned statecap_L1_demeaned xdem_L1_demeaned 1.highterror_L1 1.highdissent_L1 lngdp_L1_demeaned lnpop_L1_demeaned hhi_recip_L1_demeaned statist_donor c.statecap_L1_demeaned#c.repression_L1_demeaned i.year i.donorcode, ll(0) cl(donorcode)
est sto m14_gov_statecap
margins, dydx(repression_L1_demeaned) at(statecap_L1_demeaned=(-2(1)2))
margins, dydx(repression_L1_demeaned) at(statecap_L1_demeaned=($statecap_lo $statecap_avg $statecap_hi)) pwcompare(effects)
margins, at(repression_L1_demeaned=(-4(1)4) statecap_L1_demeaned=($statecap_lo $statecap_avg $statecap_hi))
marginsplot, legend(position(6))
graph save Graph "graphs\14_gov_statecap.gph", replace* Model 15: Gov High Terror
tobit lngovaidpc repression_L1_demeaned donorrights_L1_demeaned idealptdist_L1_demeaned lnexport_L1_demeaned statecap_L1_demeaned xdem_L1_demeaned 1.highterror_L1 1.highdissent_L1 lngdp_L1_demeaned lnpop_L1_demeaned hhi_recip_L1_demeaned statist_donor 1.highterror_L1#c.repression_L1_demeaned i.year i.donorcode, ll(0) cl(donorcode)
est sto m15_gov_terror
margins, dydx(i.highterror_L1) at(repression_L1_demeaned=(-4(1)4))
margins i.highterror_L1, at(repression_L1_demeaned=(-4(1)4))
marginsplot, legend(position(6))* Model 16: Gov Shaming
tobit lngovaidpc repression_L1_demeaned donorrights_L1_demeaned idealptdist_L1_demeaned lnexport_L1_demeaned statecap_L1_demeaned xdem_L1_demeaned 1.highterror_L1 1.highdissent_L1 lngdp_L1_demeaned lnpop_L1_demeaned hhi_recip_L1_demeaned statist_donor shame_physint_L1_demeaned c.shame_physint_L1_demeaned#c.repression_L1_demeaned i.year i.donorcode, ll(0) cl(donorcode)
est sto m16_gov_shaming
margins, dydx(repression_L1_demeaned) at(shame_physint_L1_demeaned=(-2(1)2))
margins, dydx(repression_L1_demeaned) at(shame_physint_L1_demeaned=($shame_lo $shame_avg $shame_hi)) pwcompare(effects)
margins, at(repression_L1_demeaned=(-4(1)4) shame_physint_L1_demeaned=($shame_lo $shame_avg $shame_hi))
marginsplot, legend(position(6))
graph save Graph "graphs\16_gov_shaming.gph", replace*******************************
* TABLE: GOVERNANCE SECTOR DV *
*******************************
global govtobit m9_gov_base m10_gov_donorrights m11_gov_idealpt m12_gov_export m13_gov_execconst m14_gov_statecap m15_gov_terror m16_gov_shaming
outreg2 [${govtobit}] using "tables\gov_table.xls", label replace excel